Preparation of Fabs and cryo-EM of labeled capsids
Plasma was drawn by plasmopheresis from a 43 year old Asian male with untreated active chronic Hepatitis B. The polyclonal HBcAg-specific antibodies were isolated and Fabs prepared from them (see Materials & Methods and ). This material was incubated with recombinant Cp149 capsids (i.e. capsids produced by expressing a construct for residues 1 – 149, the core domain, in E. coli) at an equimolar ratio of Fabs to Cp149 subunits. After labeling was confirmed by negative staining EM (data not shown), cryo-EM was performed. On comparing the resulting images with those of an unlabeled control (cf. ), it is evident that a dense labeling was achieved.
Cryo-electron micrographs of (A) control HBV-Cp149 capsids and (B) HBV-Cp149 capsids decorated with polyclonal human Fabs. Black arrowhead: a T=4 capsid; white arrowhead: a T=3 capsid. Scale bar: 50 nm.
Image reconstruction of Fab-labeled capsids
To characterize the binding of these antibodies in greater detail, three-dimensional density maps of the labeled capsids were calculated. The reconstruction of the T=4 capsid is shown in . Its resolution was assessed as ~ 11 Å by the Fourier shell correlation criterion with a conservative threshold of 0.5 (see Methods). A local resolution calculation (Cardone et al., Ms in preparation) indicated a value of ~ 8 Å for the capsid shell and 8 – 14 Å for the Fab-related density, the resolution being highest near the capsid-Fab interface and progressively lower, further away. The T=3 capsid map revealed similar features but was noisier and had lower resolution (data not shown). Accordingly, we focused our detailed analysis on the T=4 capsid which, moreover, is the predominant form produced in infected human livers (Kenney et al., 1995
; Dryden et al., 2006
Figure 3 (A) Model of the surface lattice of the T=4 capsid with the four quasi-equivalent subunits distinguished by different colors (green, yellow, blue, red), viewed along an axis of 2-fold symmetry (see Belnap et al., 2003). (B) Surface rendering of the Fab-decorated (more ...)
The reconstruction of the T=4 capsid reveals not only major elements of secondary structure in the shell, for example, the four α-helices in the spike bundle, but also the surrounding Fab-related density. A surface rendering of the labeled capsid is shown in with a model of the T=4 capsid shown alongside for reference (). Various features of Fab-related density are shown in . These features do not have the shapes of individual Fab molecules because they represent the superposition of Fabs bound at adjacent epitopes which cannot be simultaneously occupied because of steric hindrance. Consequently, their molecular envelopes merge in the density maps. For this reason also, i.e. substoichiometric occupancy, the Fab-related density is generally lower than in the capsid shell (). Two color codlings are used in this Figure: one, in , distinguishes the four quasi-equivalent subunits in conventional colors. The other, in , distinguishes certain features of Fab-related density.
To interpret the Fab-labeled map, we relied mainly on grayscale sections. The central section of a view along a 2-fold symmetry axis is particularly informative as it includes longitudinal and transverse sections through capsid protein dimers and it contains all three symmetry axes (5-, 3-, and 2-fold). This section () directly shows density overlying the 2-fold axis (the axis is marked in ) coming from floor-binding Fabs, and other density surrounding the spikes, corresponding to spike-binders.
Modeling of antibody structures into the cryo-EM density map
To localize the predominant epitopes, we performed modeling in which high resolution structures of the capsid and a representative Fab molecule were fitted into the labeled density map. Although the labeling mixture was expected to be highly polyclonal, the differentiated nature of the Fab density led us to infer that it was dominated by a limited number of contributors. For modeling purposes, we tried a few Fab crystal structures of different types (IgG1, IgG2, etc) and with differences in their elbow angle and fitted them into the EM density at the 5-fold of the T=4 map. The one that best reproduced this experimental density was an IgG1 kappa antibody (PDB code 1VGE). Because we were chary about introducing too many degrees of freedom into the modeling process (e.g. different elbow angles for each Fab), we used the same molecule to model all Fab-related densities.
To assess models of the labeled capsid, we compared their central sections with that of the experimental map after band-limiting the model to (14 Å)−1. At each step, the modeling was refined to optimize the match. The occupancies of the modeled Fabs at the different sites () were estimated in the same way. We started by modeling the most prominent densities – those at the AA site at the 5-fold axis - and when they were well accounted for, addressed the next most prominent densities, and so on. This procedure reached a conclusion when Fabs had been modeled as occupying five distinct epitopes that we named he1 - he5 (he = human epitope). Overall, there is non-zero occupancy of 12 of the 20 quasi-equivalent sites (5 sets of 4 each) - ). At that point, all significant densities present in the reconstruction were accounted for (cf. ). The positions of the five epitopes in question are marked on the cryo-EM density map in . Their positions on the primary sequence are mapped in and on the capsid surface in .
Summary of epitope locations and occupancies.
FIGURE 4 (A) Enlargement of a region of capsid surrounding the 3-fold axis with epitopes he1 to he5 marked in different colors. (B) Mapping of the positions of the epitope-forming peptides on the core domain sequence. (C) Mapping of the positions of the epitope-forming (more ...)
Epitopes, shown on the surface-rendered reconstruction in , are here marked on ribbon diagrams of the corresponding subunits. The five epitopes, the 2-, 3- and 5-fold symmetry axes, and the AB and CD dimers, are indicated.